fix(resm): 修复 openalex 链失败遮蔽 PDF 兜底下载

send_download_fulltext_task 原用 fail_reason=None 选取, 导致被 openalex
保活链写过 fail_reason 的论文被永久遮蔽, oa_url/elsevier/scihub 兜底路径
永不尝试。改为 download_pdf 终态打稳定标记 download_pdf_tried, 选取时据此
排除 —— 既解除遮蔽, 又防本链路对同一篇无限重试。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

apps/resm/tasks.py

@@ -712,9 +712,12 @@ def can_send_more(max_running):
 
 @shared_task(base=CustomTask)
 def send_download_fulltext_task(number_of_task=100):
-    qs = Paper.objects.filter(has_fulltext=False, fail_reason=None, is_oa=True).exclude(
+    # 只排除"本下载链路已尝试过"的论文(download_pdf 终态会打 download_pdf_tried 标记),
+    # 不再用 fail_reason=None —— 否则被 openalex 保活链失败标记蹭上 fail_reason 的论文会被
+    # 永久遮蔽, 其 oa_url/elsevier/scihub 兜底路径永远不会被尝试。
+    qs = Paper.objects.filter(has_fulltext=False, is_oa=True).exclude(
         fetch_status='downloading'
-    )
+    ).exclude(fail_reason__contains="download_pdf_tried")
     if not qs.exists():
         return "done"
     qs0 = qs.order_by("?")
@@ -772,8 +775,12 @@ def download_pdf(paper_id):
         # if paper.has_fulltext_pdf is False and cache.get("openalex_api_exceed") is None:
         #     current_from = "openalex"
         #     msg = save_pdf_from_openalex(paper)
-        if paper.fail_reason is None and paper.has_fulltext_pdf is False:
+        if paper.has_fulltext_pdf is False:
-            paper.save_fail_reason(msg)
+            # 终态标记: 无论该论文之前是否已被 openalex 保活链写过 fail_reason, 这里都追加
+            # 一条带 download_pdf_tried 的终态, 供 send_download_fulltext_task 据此排除。
+            # 既防本下载链路(oa_url/elsevier/scihub)对同一篇无限重试, 也避免 openalex 链
+            # 留下的 fail_reason 把本链路的首次尝试给遮蔽掉。
+            paper.save_fail_reason(f"download_pdf_tried:{msg}")
         return msg, current_from
     finally:
         paper.fetch_end()